Institution
Amazon.com
Company•Seattle, Washington, United States•
About: Amazon.com is a company organization based out in Seattle, Washington, United States. It is known for research contribution in the topics: Computer science & Service (business). The organization has 13363 authors who have published 17317 publications receiving 266589 citations.
Topics: Computer science, Service (business), Service provider, Context (language use), Virtual machine
Papers published on a yearly basis
Papers
More filters
••
15 Dec 2017TL;DR: This paper proposed a unified neural network that jointly performs domain, intent, and slot predictions for multi-task learning and achieved significant improvements in all three tasks across all domains over strong baselines.
Abstract: In practice, most spoken language understanding systems process user input in a pipelined manner; first domain is predicted, then intent and semantic slots are inferred according to the semantic frames of the predicted domain. The pipeline approach, however, has some disadvantages: error propagation and lack of information sharing. To address these issues, we present a unified neural network that jointly performs domain, intent, and slot predictions. Our approach adopts a principled architecture for multitask learning to fold in the state-of-the-art models for each task. With a few more ingredients, e.g. orthography-sensitive input encoding and curriculum training, our model delivered significant improvements in all three tasks across all domains over strong baselines, including one using oracle prediction for domain detection, on real user data of a commercial personal assistant.
84 citations
•
TL;DR: This paper explores effective Transformer architectures for language model, including adding additional LSTM layers to better capture the sequential context while still keeping the computation efficient, and proposes Coordinate Architecture Search (CAS) to find an effective architecture through iterative refinement of the model.
Abstract: The Transformer architecture is superior to RNN-based models in computational efficiency. Recently, GPT and BERT demonstrate the efficacy of Transformer models on various NLP tasks using pre-trained language models on large-scale corpora. Surprisingly, these Transformer architectures are suboptimal for language model itself. Neither self-attention nor the positional encoding in the Transformer is able to efficiently incorporate the word-level sequential context crucial to language modeling.
In this paper, we explore effective Transformer architectures for language model, including adding additional LSTM layers to better capture the sequential context while still keeping the computation efficient. We propose Coordinate Architecture Search (CAS) to find an effective architecture through iterative refinement of the model. Experimental results on the PTB, WikiText-2, and WikiText-103 show that CAS achieves perplexities between 20.42 and 34.11 on all problems, i.e. on average an improvement of 12.0 perplexity units compared to state-of-the-art LSTMs. The source code is publicly available.
84 citations
••
TL;DR: The potential for a statistical model, previously developed for Peruvian sunlit leaves, to consistently predict leaf ages from leaf reflectance spectra across two contrasting forests in Peru and Brazil is explored and a more general model is achieved that well-predicted leaf age across forests and environments.
Abstract: Leaf age structures the phenology and development of plants, as well as the evolution of leaf traits over life histories. However, a general method for efficiently estimating leaf age across forests and canopy environments is lacking. Here, we explored the potential for a statistical model, previously developed for Peruvian sunlit leaves, to consistently predict leaf ages from leaf reflectance spectra across two contrasting forests in Peru and Brazil and across diverse canopy environments. The model performed well for independent Brazilian sunlit and shade canopy leaves (R2 = 0.75-0.78), suggesting that canopy leaves (and their associated spectra) follow constrained developmental trajectories even in contrasting forests. The model did not perform as well for mid-canopy and understory leaves (R2 = 0.27-0.29), because leaves in different environments have distinct traits and trait developmental trajectories. When we accounted for distinct environment-trait linkages - either by explicitly including traits and environments in the model, or, even better, by re-parameterizing the spectra-only model to implicitly capture distinct trait-trajectories in different environments - we achieved a more general model that well-predicted leaf age across forests and environments (R2 = 0.79). Fundamental rules, linked to leaf environments, constrain the development of leaf traits and allow for general prediction of leaf age from spectra across species, sites and canopy environments.
84 citations
•
28 Aug 2013TL;DR: In this article, a client request for content is returned that includes executable code for generating a request for preload information, based on processing the executable code, a client computing device requests pre-load information from a content delivery service provider.
Abstract: A system, method and computer-readable medium for client-side cache management are provided. A client request for content is returned that includes executable code for generating a request for preload information. Based on processing the executable code, a client computing device requests preload information from a content delivery service provider. The content delivery service provider provides an identification of content based on resource requests previously served by the content delivery service provider. The client computing device processes the preload information and generates and obtains identified resources for maintenance in a client computing device memory, such as cache.
84 citations
•
08 Aug 2012TL;DR: In this paper, a storage service may store an index associated with archived data and update the retrieved portion of the index with at least part of the received information, based at least in part on the subsets.
Abstract: Embodiments of the present disclosure are directed to, among other things, managing inventory indexing of one or more data storage devices. In some examples, a storage service may store an index associated with archived data. Additionally, the storage service may receive information associated with an operation performed on the archived data. The storage service may also partition the received information into subsets corresponding to an identifier. In some cases, the identifier may be received with or otherwise be part of the received information. The storage service may also retrieve at least a portion of the index that corresponds to the subset. Further, the storage service may update the retrieved portion of the index with at least part of the received information. The updating may be based at least in part on the subsets.
84 citations
Authors
Showing all 13498 results
Name | H-index | Papers | Citations |
---|---|---|---|
Jiawei Han | 168 | 1233 | 143427 |
Bernhard Schölkopf | 148 | 1092 | 149492 |
Christos Faloutsos | 127 | 789 | 77746 |
Alexander J. Smola | 122 | 434 | 110222 |
Rama Chellappa | 120 | 1031 | 62865 |
William F. Laurance | 118 | 470 | 56464 |
Andrew McCallum | 113 | 472 | 78240 |
Michael J. Black | 112 | 429 | 51810 |
David Heckerman | 109 | 483 | 62668 |
Larry S. Davis | 107 | 693 | 49714 |
Chris M. Wood | 102 | 795 | 43076 |
Pietro Perona | 102 | 414 | 94870 |
Guido W. Imbens | 97 | 352 | 64430 |
W. Bruce Croft | 97 | 426 | 39918 |
Chunhua Shen | 93 | 681 | 37468 |